PulseAugur
实时 11:37:48
English(EN) Microsoft Research Releases Webwright: A Terminal-Native Web Agent Framework That Scores 60.1% on Odysseys, Up from Base GPT-5.4’s 33.5%

Microsoft Research 的 Webwright 提升了 AI Web Agent 的性能

Microsoft Research 开发了 Webwright,这是一个开源框架,允许 AI Agent 通过基于终端的方法与 Web 进行交互。与一次在一个浏览器中执行一步操作的传统 Agent 不同,Webwright Agent 在终端环境中编写和执行 Playwright 代码、bash 命令并检查日志。这种方法显著提高了性能,在 Odysseys 基准测试中取得了 60.1% 的成绩,远高于使用传统基于截图的 Agent 设置的基础 GPT-5.4 模型得分 33.5%。 AI

影响 通过采用以代码为中心的方法,使 AI Agent 能够更有效地执行复杂的 Web 任务,从而可能提高自动化和数据提取能力。

排序理由 该集群描述了 Microsoft Research 发布的一个用于 AI Agent 的新开源框架,包括基准测试结果。

在 MarkTechPost 阅读 →

AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →

Microsoft Research 的 Webwright 提升了 AI Web Agent 的性能

报道来源 [3]

  1. MarkTechPost TIER_1 English(EN) · Asif Razzaq ·

    Microsoft Research Releases Webwright: A Terminal-Native Web Agent Framework That Scores 60.1% on Odysseys, Up from Base GPT-5.4’s 33.5%

    <p>Microsoft Research introduces Webwright, a terminal-native browser agent framework that replaces click-trace web automation with reusable Playwright scripts. Using a single agent loop across three modules and roughly 1,000 lines of code, Webwright powered by GPT-5.4 reaches 60…

  2. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    https:// winbuzzer.com/2026/05/25/micro soft-webwright-turns-web-agents-into-reusable-code-xcxwbn/ Microsoft Research has released Webwright, a Playwright scrip

    https:// winbuzzer.com/2026/05/25/micro soft-webwright-turns-web-agents-into-reusable-code-xcxwbn/ Microsoft Research has released Webwright, a Playwright script-first web-agent framework that moves persistent state into a terminal workspace. # AI # Webwright # MicrosoftResearch …

  3. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Microsoft Research has unveiled Webwright, a terminal-native web agent framework that achieves 60.1% on the Odysseys benchmark, a significant leap from the base

    Microsoft Research has unveiled Webwright, a terminal-native web agent framework that achieves 60.1% on the Odysseys benchmark, a significant leap from the base GPT-5.4 score of 33.5%. The framework enables autonomous AI agents to navigate the web independently. https://www. mark…